AI clocked in, generated their own ads, and Houdini’d their way out of a locked box.

First, they came for the side hustle: Brands and TikTok’s algorithm are leaning into AI-generated creators, squeezing out humans who built businesses on shoppable videos.

GMV Max is actively surfacing these synthetic “little billboards,” running the same flash-sale scripts that real humans would. Only for a fraction of the cost. Uh-oh.

Then they moved into advertising: AI assistants are becoming a real ad channel now with the numbers confirming it:

• ChatGPT is testing two ads from two advertisers in its footer, a sign inventory is opening up.
• Meanwhile, Google AI Mode served ads on almost a third of commercial keywords tested. But paid slots matched citations only 11% of the time.

Translation: buying a slot doesn’t buy you a mention. Track them separately.

Also, ChatGPT’s share of AI users dipped below 50% for the first time as Gemini and Claude gained. Which means audiences are fragmenting, and so should your AI ads strategy.

But here’s the catch: An OpenAI model recently chained zero-days to breach live infrastructure during a security eval.

The tools are getting more useful and more ad-friendly, but the security unknowns are still very real. How deeply you wire AI into your workflow should account for that.